一种等概率分布式检测系统的反馈自适应学习算法

A Feedback Adaptive Algorithm for Decentralized Detection System

研究了分布式检测系统在等概率情况下的最优检测问题。针对传感器虚警与漏报概率未知的情况,提出了一种状态反馈自适应学习算法,通过在线的修正融合权值,最终使系统收敛于最佳权值,并对算法收敛性和误差进行了理论分析。最后给出的仿真算例证实了理论结果。

Most papers by past researchers on decentralized multi-sensor detection system, usually employing fixed-value fusion weight coefficients, appear unable to keep the system in optimized detection status when the detection probability is unknown or varying. We propose a feedback adaptive learning algorithm to meet the optimized detection requirement. In the fusion scheme of the adaptive algorithm proposed by us, information system can estimate the Bayes fusion weight coefficients online. In the full paper, we explain in much detail the adaptive algorithm proposed by us; here we just list the three topics discussed in our detailed explanation： （1） adaptive algorithm; （2） analysis of convergence of fusion weight coefficients; one important result is that, under certain conditions, the fusion weight coefficients will converge to their optimized values; （3） error analysis. Finally we give a numerical simulation examplet the variations of fusion weight coefficients with number of iterations are shown in Fig. 1 of the full paper. Fig. 1 shows that almost all fusion weight coefficients converge to their optimized values after about 1500 iterations.